STAGE-Claw: leaderboard
STAGE-Claw (2026) builds 40 realistic personal-agent tasks automatically and scores an agent in the OpenClaw framework by the verified final state of real calendars, mail, reminders and files rather than by its text reply, with reasoning disabled where a model allows it. The board is the mean task score from 0 to 100 given by the verification programs; higher is better.
Metric: Average task score (0-100) from the verification programs, 40 automatically built and human-audited personal-agent tasks run in real operating-system environments (calendar, email, reminders, files); the final system state is checked by task-specific verification programs; OpenClaw agent framework, reasoning disabled where the model allows it, temperature 0 where configurable, first valid run per task; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 77.1 | #79 |
| 2 | Claude Sonnet 4.6 | 69.43 | #161 |
| 3 | GPT-5.5 (Non-reasoning) | 69.19 | #49 (GPT-5.5) |
| 4 | GLM-5 (Non-reasoning) | 67.4 | #260 (GLM-5) |
| 5 | Gemini 3.1 Pro (Preview) | 65.5 | #94 |
| 6 | GPT-5.4 (Non-reasoning) | 65.37 | #123 (GPT-5.4) |
| 7 | DeepSeek V4 Pro (Non-reasoning) | 59.78 | #188 (DeepSeek V4 Pro) |
| 8 | Kimi K2.6 (Non-reasoning) | 59.05 | #178 (Kimi K2.6) |
| 9 | MiniMax-M2.7 | 47.53 | #518 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=stage-claw · How It Works · Data refreshed daily, snapshot 2026-11-13.